PRE-DEPLOYMENT ETHICS REVIEW GDPR COMPLIANT EU AI ACT ALIGNED 100% EUROPEAN INFRASTRUCTURE
Pilot Research Proposal Pre-Deployment Phase Ethics Review Submitted Confidential
Research · Pilot Study Proposal

Research Proposal

Conversational AI for Behavioural Recovery Support. A pilot research study proposal submitted voluntarily for ethics review — prior to any real-world deployment. Seeking institutional co-investigators across Europe and the United Kingdom.

Applicant Francisco Moreno · Anborela OÜ
Document status Submitted for ethics review · Pre-deployment
Contact ai@anborela.ee · Estonia / Spain
01

Project Summary

This project involves the development of an experimental conversational artificial intelligence system designed to provide supportive dialogue to people in recovery situations, including emotional distress, addiction recovery, and behavioural self-regulation.

The system is currently in a training and testing phase and is not publicly deployed. The objective of this pilot study is to evaluate whether a conversational AI model can provide safe, supportive, and non-clinical dialogue — without replacing professional treatment. The system is conceived exclusively as a research prototype.

Core principle

Not a treatment. A research instrument.

The ANBORELA system does not constitute medical or psychological treatment. It is designed as a complement to — never a substitute for — professional clinical care. Every interaction includes a visible reminder of this boundary and emergency referral information.

02

Purpose of the Study

This pilot study has four specific research objectives:

Evaluate the behaviour of a conversational AI model in recovery support situations Analyse user interaction patterns across different recovery stages Identify potential risks in AI-based dialogue systems operating in health-adjacent domains Develop evidence-based safety guidelines for conversational AI in sensitive contexts

This study does not aim to provide medical or psychological treatment.

03

AI System Description

The ANBORELA system is a text-based conversational model trained using the following data sources:

Public linguistic corpora Academic text sources Manually created dialogue examples Synthetic training conversations General language models

No personal medical records are used at any stage. The system generates responses automatically from user input, does not perform diagnoses, and includes hardcoded safety rules to prevent harmful responses.

Access channel Telegram · No app installation required beyond Telegram itself
Infrastructure 100% European servers · Independent from AWS, Azure and Google Cloud
Availability 24/7 · Fully anonymous · No personal data linked to identity
Clinical scope Non-clinical · Does not diagnose, prescribe or replace professional care
04

Current Project Status

Development phase Training and testing in a controlled environment
Real users None involved at this stage
Public access Not available — ethics approval required before any real-world testing
Ethics review Voluntarily submitted · Seeking national and university committee approval
05

Data Collection

During the pilot phase, the following data may be collected:

Text conversations Anonymous interaction logs Timestamps and session duration System responses for safety evaluation

No personal identification is required. Users will not be asked for their real name, identification number or medical history. All data will be stored securely on European infrastructure.

06

Data Protection

All data processing follows GDPR principles. Specific measures include:

Data anonymised at the point of collection Secure encrypted storage on European servers No data sold to third parties under any circumstances No transfer to third parties without explicit authorisation Data used exclusively for research purposes

The system will include informed consent mechanisms before participation begins. Users may withdraw from the study at any time without consequence.

07

Risk Assessment

Risk Emotional distress · Clear notice displayed to all users before interaction begins. Emergency services information always visible and cannot be disabled.
Risk Misinterpretation of AI responses · The system explicitly states it does not constitute medical advice at every interaction. Non-clinical scope is hardcoded, not configurable.
Risk Excessive dependency on the system · Dialogue scope is limited and controlled. The system actively encourages users to seek professional support. Option to stop the conversation is always available.

The system is expressly designed not to replace professional clinical care.

08

Ethical Considerations

The project is voluntarily submitted for ethics evaluation before any public use. The developer seeks guidance from ethics committees to ensure compliance with European research regulations and standards.

The objective is to create a safe experimental framework for research into conversational AI in sensitive domains adjacent to health — contributing to the evidence base for responsible AI development in Europe.

Lived experience · Francisco Moreno

The differentiator no institution can replicate

The project is led by Francisco Moreno, who brings 30+ years of lived experience with alcohol and cocaine addiction, treated through Spain's UCA clinical units. This perspective — from inside the system as a patient, survivor, and now innovator — is the project's most powerful differentiator for European funding bodies and ethics committees. It ensures every design decision is grounded in the reality of what people in recovery actually experience.

09

Expected Outcomes

This study aims to generate the following outputs for the European research community:

Evaluation of AI behaviour in recovery dialogue contexts Identification and documentation of risks in health-adjacent conversational AI Evidence-based safety guidelines for conversational AI in sensitive domains Foundation for future academic collaborations and Horizon Europe applications
10

Contact and Partnership

For ethics review requests, co-investigator enquiries, or institutional partnership discussions:

Name Francisco Moreno · Founder & DPO
Organisation Anborela OÜ · Registration 14746683 · Estonia
Address Pärnu mnt 139c, Tallinn, Estonia

Ready to co-design the evidence? We are inviting European universities and research bodies to join as institutional co-investigators.

Anborela OÜ · Registration 14746683 · Estonia · Submitted for ethics review — pre-deployment phase
Research Proposal v1.0 · May 2026 · Confidential · The ANBORELA system does not constitute medical or psychological treatment